diff --git a/examples/text-generation/quantization_config/unit_scale_quant.json b/examples/text-generation/quantization_config/unit_scale_quant.json
index 216cf27e68..20783ea3f1 100644
--- a/examples/text-generation/quantization_config/unit_scale_quant.json
+++ b/examples/text-generation/quantization_config/unit_scale_quant.json
@@ -3,5 +3,10 @@
     "mode": "QUANTIZE",
     "observer": "maxabs",
     "scale_method": "unit_scale",
-    "dump_stats_path": "./hqt_output/measure"
+    "whitelist": {"types": [], "names":  []},
+    "blacklist": {"types": [], "names":  []},
+    "quantize_weight": false,
+    "dump_stats_path": "./results/hk",
+    "ignore_modules_wo_measures": "True",
+    "dump_stats_xlsx_path": "./run_outputs/fp8stats.xlsx"
 }
diff --git a/examples/text-generation/utils.py b/examples/text-generation/utils.py
index c4bd2147ac..7a4ae91425 100644
--- a/examples/text-generation/utils.py
+++ b/examples/text-generation/utils.py
@@ -430,7 +430,12 @@ def setup_distributed_model(args, model_dtype, model_kwargs, logger):
     logger.info("DeepSpeed is enabled.")
     deepspeed.init_distributed(dist_backend="hccl")
     config = AutoConfig.from_pretrained(args.model_name_or_path, torch_dtype=model_dtype, **model_kwargs)
-    load_to_meta = model_on_meta(config)
+
+    keep_module_on_host = False
+    if "Llama-3.1-405B" in args.model_name_or_path:
+        keep_module_on_host = True
+
+    load_to_meta = False if keep_module_on_host else model_on_meta(config)
 
     if args.assistant_model is None:
         assistant_model = None
@@ -485,6 +490,7 @@ def setup_distributed_model(args, model_dtype, model_kwargs, logger):
 
     # Initialize the model
     ds_inference_kwargs = {"dtype": model_dtype}
+    ds_inference_kwargs["keep_module_on_host"] = keep_module_on_host
     ds_inference_kwargs["tensor_parallel"] = {"tp_size": args.world_size}
     ds_inference_kwargs["enable_cuda_graph"] = args.use_hpu_graphs
     ds_inference_kwargs["injection_policy"] = get_ds_injection_policy(config)